This is a problem that is pretty common: generating a relation on the fly without creating a table. SQL solutions for this problem are pretty awkward. One example using a derived table:
SELECT n.id
FROM
(SELECT 2 AS id
UNION SELECT 3
UNION SELECT 4
UNION SELECT 5
UNION SELECT 6
UNION SELECT 7) AS n
LEFT OUTER JOIN foos USING (id)
WHERE foos.id IS NULL;
But this doesn't scale very well, because you might have many values instead of just six. It can become tiresome to construct a long list with one UNION
needed per value.
Another solution is to keep a general-purpose table of ten digits on hand, and use it repeatedly for multiple purposes.
CREATE TABLE num (i int);
INSERT INTO num (i) VALUES (0), (1), (2), (3), (4), (5), (6), (7), (8), (9);
SELECT n.id
FROM
(SELECT n1.i + n10.i*10 AS id
FROM num AS n1 CROSS JOIN num AS n10
WHERE n1.i + n10.i*10 IN (2, 3, 4, 5, 6, 7)) AS n
LEFT OUTER JOIN foos USING (id)
WHERE foos.id IS NULL;
I show the inner query generating values from 0..99 even though this isn't necessary for this case. But you might have values greater than 10 in your list. The point is that with one table num
, you can generate large numbers without having to resort to very long chains with one UNION
per value. Also, you can specify the list of desired values in one place, which is more convenient and readable.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…